Relating target tracking

1. A method of tracking a target at a local node in a decentralised network having a plurality of nodes, each node comprising a processor, a sensor operable to make target observations, communication means operable to transmit and receive track information, and storage means operable to store track information at the local node, the method comprising: in the processor of each node: (i) associating each initial model probability of a set of initial model probabilities with a target dynamics model of a set of target dynamics models, wherein each initial model probability is related to a probability that target dynamics are in accordance with an associated target dynamics model; (ii) associating each initial model target state estimate of a set of initial model target state estimates with the target dynamic model of the set of target dynamics models; (iii) calculating a set of predicted model target state estimates by applying each target dynamics model to an associated initial model target state estimate; (iv) in response to a target observation being made at the local node, performing a first tracking procedure, thereby updating track information stored at the local node; (v) in response to track information being received at the local node from a remote node, performing a second target tracking procedure, thereby updating track information stored at the local node; and (vi) communicating updated track information to selected other nodes in the network; and wherein the first tracking procedure comprises the steps of: (a) calculating updated model probabilities in dependence on a difference between the target observation and the predicted model target state estimate; (b) calculating updated model target state estimates in dependence on the predicted model target state estimate and the target observation; and (c) calculating a composite predicted target state estimate by combining the predicted model target state estimates using a weighting related to the updated model probabilities; and the second tracking procedure comprises the step of conservatively fusing the received track information with track information stored at the local node to produce fused track information; and wherein the track information comprises one or more of: the updated composite target state estimate, the updated model target state estimates, and the updated model probabilities.

2. The method as claimed in claim 1 , further comprising the steps of repeating steps (i) to (vii), at each node in the network, and then iterating steps (iv) to (vii), at each node in the network by using, in place of the initial model probabilities and the initial model target state estimates, the updated model probabilities and the updated model target state estimates from the immediately preceding iteration.

3. The method as claimed in claim 2 , further comprising the step of mixing the initial model target state estimate by combining, with a first initial model target state estimate, a proportion of each of the other initial model target state estimates, the proportion being related to the probability that, during the immediately preceding iteration, the target dynamics transitioned from being in accordance with the target dynamics model associated with said other initial model target state estimate, to being in accordance with the target dynamics model associated with the first initial model target state estimate; and repeating the combination procedure for each initial model target state estimate.

4. The method as claimed in claim 3 , wherein the first tracking procedure further comprises the step of associating the target observation with track information stored at the local node.

5. The method as claimed in claim 4 , wherein the second tracking procedure further comprises the step of associating the received track information with track information stored at the local node.

6. The method as claimed in claim 5 , wherein the step of conservatively fusing the received track information with track information stored at the local node comprises applying a covariance intersection algorithm.

7. The method as claimed in claim 5 , wherein the tracking information comprises the updated composite target state estimate.

8. The method as claimed in claim 7 , wherein the plurality of nodes comprises at least a first node and a second node, and wherein the set of target dynamics models provided at the first node is different to the set of target dynamics models provided at the second node.

9. The method as claimed in claim 5 , wherein the track information comprises the updated model target state estimates and the updated model probabilities.

10. The method as claimed in claim 9 , wherein the same set of target dynamics models is provided to each node in the network.

11. The method as claimed in claim 10 , wherein the second tracking procedure comprises conservatively fusing each received updated model target state estimate with its corresponding local model target state estimate, and conservatively fusing each received updated model probability with its corresponding local probability.

12. The method as claimed in claim 5 , wherein the second tracking procedure comprises the step of updating the initial model probabilities through application of a Bayesian network.

13. A method of tracking a target, the method being applied to a decentralised network comprising a plurality of nodes, the method comprising the steps of: in a processor of a first node: (i) performing a multiple models tracking algorithm, thereby updating track information stored at the first node; (ii) communicating updated track information to selected other nodes in the network; and (iii) in response to receiving track information from another node, conservatively fusing the receiving track information with local track information; and in a processor of each other node in the network: (iv) performing a multiple models tracking algorithm to update track information stored at each respective other node; and (v) performing steps (ii) and (iii) above at each respective other node in the network.

14. The method as claimed in claim 13 , wherein the multiple models tracking algorithm is an interacting multiple models tracking algorithm.

15. A decentralised network of nodes, wherein each node comprises: sensing means for detecting tracking information of a target; means for providing each target dynamics model with a model target state estimate and a model probability; means for applying each target dynamics model to an associated initial model target state estimate; means for calculating updated model probabilities based on a difference between tracking information of a target and a predicted model target state estimate; means for calculating updated model target state estimates based on the predicted model target state estimate and the tracking information of the target; means for calculating a composite predicted target state estimate by combining the predicted model target state estimates using a weighting related to the updated model probabilities to generate updated tracking information; and communication means for communicating the updated tracking information to selected other nodes in the network.

16. A decentralized network of nodes, wherein each node comprises: sensing means for detecting tracking information of a target; storage means for storing tracking information, means for providing each target dynamics model with a model target state estimate and a model probability; means for applying each target dynamics model to an associated initial model target state estimate; means for associating tracking information received from another node with tracking information stored in said storage means; means for fusing the associated tracking information received from another node and the tracking information stored in said storage means, means for updating the tracking information stored in said storage means with the fused tracking information; and communication means for communicating the updated tracking information to selected other nodes in the network.